This was a fun read. I appreciated the recurring attention to “Deep Blue“ and laughed at this part:

> These projects were quite useful, in that they sort of cured me of my AI mania... because after I built these things, I got to look at them and ask “does the world need a slow, buggy, half-baked Python JavaScript interpreter?”

> I don’t think the world does.

Very relatable.

Reliable enough daily, still not repeatable. The same task, same config, same model cost me $2.11 one run and $1.33 the next. Run-to-run variance beat most config changes I measured.

All future whiteboard interviews, at the big five, will include a section on drawing Pelicans.